Insight. Applied.
An audit report has blamed a spreadsheet "copy and paste" error for cost overruns and delays at an Edinburgh children’s hospital.
A local politician called it
[…] one of the most expensive typos in history.
How many more high-profile spreadsheet failures is it going to take before we deem the inappropriate use of them to be professional malpractice?
When the lead story in the Daily Mail complains about Excel misuse in its headline you know things have gone too far.
15,841 coronavirus cases were excluded from UK government figures as a result of
…an Excel spreadsheet reaching its maximum file size, which stopped new names being added in an automated process.
Apparently these details were not supplied to the "track and trace" programme, meaning that people exposed to the virus were not alerted—potentially leading to unnecessary infections.
I’ve been complaining about spreadsheet misuse for a decade, and this is one of the most shambolic examples I’ve come across. Blatant abuse of Excel as a ramshackle database.
This also highlights the dangers of casual automation. Without appropriate checks and balances, automated processes can fail catastrophically, and silently.
There really is no excuse for this level of incompetency. And, the worst thing is the "solution".
The technical issue has now been resolved by splitting the Excel files into batches.
Add more sticking plaster to the homespun system and press on. I can only weep.
Some additional technical details have been revealed since the story broke.
Business Insider has an infographic that describes 20 of the most common cognitive biases that can screw up your decision-making.
Those of us involved in decision and data science would do well to remember the words attributed to John R. Searle.
If you have to add “scientific” to a field, it probably ain’t.
Statistical confidence intervals are almost always misinterpreted. Consider the following statement.
"The prevalence of the disease P has a 95% confidence interval of 1% <= P <= 5%."
This is commonly taken to imply that there’s a 95% chance that the true prevalence is between 1% and 5%.
This isn’t the case.
Confidence intervals represent uncertainty about the interval, rather than the parameter of interest.
The correct interpretation of the confidence interval defined above is that if we collect many samples from the population and calculate confidence intervals from them, 95% of those confidence intervals will contain the true value of P.
In Bayesian statistics we generally calculate credible intervals which are compatible with the intuitive interpretation.
